20231088/vllm

History

[Kernel] Update fused_moe tuning script for FP8 (#4457 )

This PR updates the tuning script for the fused_moe kernel to support FP8 and also adds configurations for TP4. Note that for the configuration I removed num_warps and num_stages for small batch sizes since that improved performance and brought the benchmarks on par with the numbers before in that regime to make sure this is a strict improvement over the status quo.

All the numbers below are for mistralai/Mixtral-8x7B-Instruct-v0.1, 1000 input and 50 output tokens.

Before this PR (with static activation scaling):

qps = 1: 9.8 ms ITL, 0.49s e2e latency
qps = 2: 9.7 ms ITL, 0.49s e2e latency 
qps = 4: 10.1 ms ITL, 0.52s e2e latency
qps = 6: 11.9 ms ITL, 0.59s e2e latency
qps = 8: 14.0 ms ITL, 0.70s e2e latency
qps = 10: 15.7 ms ITL, 0.79s e2e latency

After this PR (with static activation scaling):

qps = 1: 9.8 ms ITL, 0.49s e2e latency
qps = 2: 9.7 ms ITL, 0.49s e2e latency
qps = 4: 10.2 ms ITL, 0.53s e2e latency
qps = 6: 11.9 ms ITL, 0.59s e2e latency
qps = 8: 11.9 ms ITL, 0.59s e2e latency
qps = 10: 12.1 ms ITL, 0.61s e2e latency

2024-05-01 11:47:38 -07:00

kernels

[Kernel] Update fused_moe tuning script for FP8 (#4457 )

2024-05-01 11:47:38 -07:00

backend_request_func.py

[Bugfix] fix output parsing error for trtllm backend (#4137 )

2024-04-17 11:07:23 +00:00

benchmark_latency.py

[Bugfix] Get available quantization methods from quantization registry (#4098 )

2024-04-18 00:21:55 -07:00

benchmark_prefix_caching.py

[Core] Enable prefix caching with block manager v2 enabled (#4142 )

2024-05-01 11:20:32 -07:00

benchmark_serving.py

[Misc] Update ShareGPT Dataset Sampling in Serving Benchmark (#4279 )

2024-04-24 09:49:13 -07:00

benchmark_throughput.py

[Misc] Use public API in benchmark_throughput (#4300 )

2024-04-24 21:10:24 +00:00

launch_tgi_server.sh

Serving Benchmark Refactoring (#2433 )

2024-02-12 22:53:00 -08:00

README.md

Change the name to vLLM (#150 )

2023-06-17 03:07:40 -07:00

sonnet.txt

feat(benchmarks): Add Prefix Caching Benchmark to Serving Benchmark (#3277 )

2024-03-27 13:39:26 -07:00

README.md

Benchmarking vLLM

Downloading the ShareGPT dataset

You can download the dataset by running:

wget https://huggingface.co/datasets/anon8231489123/ShareGPT_Vicuna_unfiltered/resolve/main/ShareGPT_V3_unfiltered_cleaned_split.json